| To better analyze vibration signals, time-domain or frequency-domain analysis has been extended to time-frequency analysis. To perform a time-frequency analysis, functions of time and frequency are designed. In the mean time, energy density or intensity of signals at different time and frequency is described. Vibration signals are processed by time-frequency analysis, in order to determine instantaneous frequency and amplitude at different moments and study time-varying signals. In this paper, failures of rotary machines are diagnosed by time-frequency analysis to find out laws from simulation data, in order to explore vibration signals from the perspective of feature extraction, classification and failure diagnosis. Intelligent software is designed by embedded technologies for monitoring and analyzing rotary machines, to analyze, diagnoise and monitor vibration signals on a real-time basis. In this paper, several aspects are summarized as follows:(1) A method based on S transform and VPMCD is proposed to analyze vibration signals. By this method, time-frequency matrix is obtained via S transform, and corresponding energy is determined along the frequency axis to compose an energy vector. Then, internal relationships among characteristic values of energy vectors are found out by VPMCD. Components with the same sum of squres of predicted errors are classified as the same category. The frequency resolution isn’t enough when vibration signals analyzed by short-time Fourier transform can be overcome by disgnostic methods based on S transform. Experimental results sugges that this method is highly effective for classifying data on features of failures.(2) A method is put forward for performing chaos and fractal analysis based on generalized S transform. In processing vibration signals by S transform, the distribution is chaotic within the main time-frequency domain of energy, so it is difficult to determine reasonable limits for extracting features. Based on window functiosns and generalized S transform, major energy of signals is distributed and more centralized within the time-frequency domain. In this way, it is advantageous for regulating and changing resolution at high temporal frequency. Feature-related informaiton included in generalized S matrix is characterized according to characteristics of box-counting dimensions by chaos and fractals, so as to accurately extract features of vibration signals. After a comparison of the experimental results of S transform, the method is demonstrated to be effective.(3) A scheme is put forwaard for designing intelligent software for monitoring and analyzying rotary machines based on generalized S transform. Developed on Visual Studio 2012, the software processes gathered data with energy spectrums based on generalized S transform through inbuilt Matlab Engine, to perform a time-frequency analysis on vibration signals of rotary machines. Besides, VPMCD failure are classified and diagnosed to analyze and dignose vibration signals for the final purpose of monitoring operations of equipment. In view of problems with known algorithms of time-frequency analysis, a range of practical methods are put forward and developed for feature extraction and classification as well as failure diagnosis. Relatively good effects are demonstrated by data simulation for roller bearings.So the methods proposed in this paper are proven to be accurate and feasible. In the mean time, monotoring and analysis software based on research methods of dissertations is designed by exploring software design with embedded technologies. In addition, data are acquired by a vibration signal collector developed in the laboratory for tesing from motors of centrifugal pumps under practical laboratory environment.
|
PDF Full Text Request